VectorVelocity

VectorVelocity is a fun and interactive tool that gamifies the comparison of different AI language models' speeds. It visualizes the token generation speed of various AI providers in a racing game format.

What it does

• Compares the speed of different AI models in generating text
• Visualizes the comparison as a car race
• Allows users to input custom prompts
• Supports multiple AI providers (Groq, SambaNova, NVIDIA, TogetherAI and more coming soon and you can also add your own!)
• Provides options for different race modes (instant or store-and-race)

DEMO

• YouTube Enjoy the videos and music you love, upload original content, and share it all with friends, family, and the world on YouTube.

Setup

1. Clone the repository:

   git clone https://github.com/BarathwajAnandan/VectorVelocity.git
   

2. Install dependencies:

   npm install
   

3. Set up your environment variables: Create a .env file in the root directory and add your API keys:

   GROQ_API_KEY=your_groq_api_key
   SNOVA_API_KEY=your_sambanova_api_key
   TOGETHERAI_API_KEY=your_togetherai_api_key
   ....
   

4. Run the application:

   npm start
   

Who is it for?

VectorVelocity is designed for:

• AI enthusiasts who want to compare different language models
• Developers looking to choose the fastest AI provider for their projects
• Curious minds eager to explore AI model performance through an fun racing game.
• Anyone who wants to turn dry performance metrics into an adrenaline-pumping competition.

Note

This is an early-stage project, and the code is still in development. Contributions, suggestions, and feedback are welcome!

Adding Custom Endpoints or Providers

VectorVelocity is designed to be easily extensible. You can add your own endpoints or other providers (like AWS, Replicate, or self-hosted models) by following these steps:

1. Modify the api_providers.json file: Add a new entry for your provider, following the existing structure. For example:

   "YOUR_PROVIDER": {
     "name": "Your Provider Name",
     "apiUrl": "https://your-api-endpoint.com/v1/chat/completions",
     "apiKey": "process.env.YOUR_PROVIDER_API_KEY",
     "models": [
       "model-name-1",
       "model-name-2"
     ],
     "selectedModel": "model-name-1",
     "makeApiCall": "makeYourProviderApiCall",
     "tokensPerSecondList": []
   }

2. Add the API key to your .env file if needed:

   YOUR_PROVIDER_API_KEY=your_api_key_here
   

3. Create an API call wrapper function in calculate_tokens.js: Add a new function similar to the existing ones, like this:

   async function makeYourProviderApiCall(prompt) {
     const headers = {
       'Authorization': `Bearer ${API_PROVIDERS.YOUR_PROVIDER.apiKey}`,
       'Content-Type': 'application/json'
     };
     
     const payload = {
       model: API_PROVIDERS.YOUR_PROVIDER.selectedModel,
       messages: [
         { role: 'system', content: 'You are a helpful assistant' },
         { role: 'user', content: prompt }
       ],
       stream: true
       // Add any other required parameters for your API
     };
   
     const response = await fetch(API_PROVIDERS.YOUR_PROVIDER.apiUrl, {
       method: 'POST',
       headers: headers,
       body: JSON.stringify(payload)
     });
   
     if (!response.ok) {
       throw new Error(`HTTP error! status: ${response.status}`);
     }
   
     return response.body.getReader();
   }

4. The application will automatically pick up the new provider from the api_providers.json file and include it in the race.

Contributing

We welcome contributions to VectorVelocity! If you have suggestions for improvements or encounter any issues, please feel free to open an issue or submit a pull request.

⭐ If you like VectorVelocity, don't forget to star the repository!

License

This project is licensed under the MIT License - see the LICENSE file for details.